R help - Jun 2007 - two basic question regarding model selection in GAM

Qusetion #1
*********  
Model selection in GAM can be done by using:
1. step.gam {gam} : A directional stepwise search
2. gam {mgcv} : Smoothness estimation using GCV or UBRE/AIC criterion

Suppose my model starts with a additive model (linear part + spline part).
Using gam() {mgcv} i got estimated degrees of freedom(edf) for the smoothing
splines. Now I want to use the functional form of my model taking estimated
degrees of freedoms in step.gam() {gam} to search a better model.

You know mgcv masks over gam. So i can not use gam after using mgcv. Is
there any way to stop mgcv.

Qusetion #2
*********
Suppose i have three models:
M1. GAM with thin plate regression spline(TPRS)
M2. GAM with cubic regression spline(CRS)
M3. GAM with some TPRS and CRS

To choose best model among the three, can i use their GCV/AIC/UBRE
criterion?
-- 
View this message in context:
http://www.nabble.com/two-basic-question-regarding-model-selection-in-GAM-tf3963362.html#a11248016
Sent from the R help mailing list archive at Nabble.com.

On Friday 22 June 2007 09:18, spime wrote:
> Qusetion #1
> *********
> Model selection in GAM can be done by using:
> 1. step.gam {gam} : A directional stepwise search
> 2. gam {mgcv} : Smoothness estimation using GCV or UBRE/AIC criterion
>
> Suppose my model starts with a additive model (linear part + spline part).
> Using gam() {mgcv} i got estimated degrees of freedom(edf) for the
> smoothing splines. Now I want to use the functional form of my model taking
> estimated degrees of freedoms in step.gam() {gam} to search a better model.
>
> You know mgcv masks over gam. So i can not use gam after using mgcv. Is
> there any way to stop mgcv.
detach(package:mgcv)
>
> Qusetion #2
> *********
> Suppose i have three models:
> M1. GAM with thin plate regression spline(TPRS)
> M2. GAM with cubic regression spline(CRS)
> M3. GAM with some TPRS and CRS
>
> To choose best model among the three, can i use their GCV/AIC/UBRE
> criterion?
Yes (assuming you're not using neg bin with unknown theta). But are the
models
very different?

simon
-- 
> Simon Wood, Mathematical Sciences, University of Bath, Bath, BA2 7AY UK
> +44 1225 386603  www.maths.bath.ac.uk/~sw283